| ▲ | retwakm4 3 hours ago | |||||||||||||
I think City governments don't have the scale for it to be worth it to the same extent. What I'm about to fantasize about is a "bucket list coding project" type thing, but I'd be happy to take a look at the famous IRS tax base mainframe code though. I'm a mainframe systems programmer and we hear stories about the "master file" and oodles of assembly code, which is right up my alley. Could be transformed to C or C++ today for easier maintainability, but my guess is that they are maybe running without LE runtime services? So maybe that's why it's a challenge to upgrade? Still, it's possible, if you manually initialize libc with an assembler macro and setup your own prelude/epilogue for initialization, comedown, abort cleanups, register passing conventions, and put assembly thunks for all the places you are needing 24-bit memory etc or accessing certain parts of OS common state... For those not in the know, on z/OS, LE (short for language environment, they didn't call it libc even though that's basically what it is because a lot of other mainframe compilers use it too and the dominance of C at the time on the platform was far from assumed for systems code, but is now probably the case) is basically "libc" (and specifically, here I'll be talking about the crt0 part). Without it you need to make your own allocators, handle all your own register save areas, etc. Back in the day when the IRS first computerized tax, the concept of a system wide "libc" you would link with didn't really exist. Every language had its own runtime library, and a lot of shops still wrote everything like it was a bare metal program using "OS" linkage which is a euphemism for basically linking the way the OS code does, not really "userspace" linkage like modern LE enabled XPLINK (much closer to the ABI and calling convention on a platform like x86_64 Linux, and much more efficient too in terms of padding arguments in registers) is. Nowadays, you want to make sure the C runtime initializes first so it can clean up your messes (like, stuff that the runtime allocates during an uncontrolled failure), even if you are programming in assembly, and if not, at least you need recovery routines to jump to. Not always possible if you are accessing certain types of OS control blocks or low storage or addressing physically (running in DAT off part of the "nucleus" aka kernel), but still, would love to see the famous file. And yes, the moment I saw who DOGE was bringing to the IRS to deal with this issue and that it was some hipster web dev groupie, I knew it was gonna flop. Probably ran away in terror when they saw what it was REALLY like programming in the 60s and the 70s as far as perf consciousness went. And that it always universally meant assembly (in a very historic and long lived, but also very CISC-y, ISA, although I suppose it was less CISC-y at the time) because compilers weren't great at optimizing yet. Anyway, from my position of expertise, that's always been the one governmental piece of code that I thought would be on my bucket list. Who knows, I'll probably end up seeing it one day anyway if I continue consulting. Lots of consultants at the IRS needed to keep that same logic churning....but with more people who knows what you could do...if all the systems software engineers in the US were contributing you could probably get it into C or C++ using a modern version of either of those languages (the latest compiler is I think clang 22 on the platform, possibly even clang 23 for those who compiled themselves or are on a beta branch, so far from an ancient C compiler that doesn't support modern C/C++ standards, and far from the ones back in the days when the IRS code was being written that couldn't optimize well enough to fit in the much more limited memory and CPU budgets of the time.) Such a task would be a Herculean effort ideally suited for an open source scenario, because you probably NEED every single person in the country that happens to know mainframe assembly and C or C++ if you want it to end up being a success. Especially in the assembly area, there really aren't that many of us, and we are spread out across all the big companies all over the country and can't coordinate a project like that so long as it's closed source. | ||||||||||||||
| ▲ | fn-mote 2 hours ago | parent [-] | |||||||||||||
> I'll probably end up seeing it one day anyway if I continue consulting. Where are all of those hackers when you need them? Too bad this the tapes (punch cards? lol) containing that software aren't leaked. | ||||||||||||||
| ||||||||||||||